In a cooperative communication system, the DECODER is replaced by a cooperative decoder, often a substantially more complex device with additional storage, multiple soft-symbol interfaces, and, in some instances, a more complex interface to the network layer. The actions of the cooperative decoder depend on whether a node plays the role of an intermediate relay or a final destination of a cooperative link. A cooperative PHY layer implementation must support both roles in a single device. We next consider the role of a final destination and describe issues related to a relay role in Section 6.5.2.
Cooperative PHY Layer: Destination Node Receiver
Consider the orthogonal relaying strategies described in Sections 5.4 and 5.5. Such strategies perform packet combining, i.e., they decode a message using the possibly unreliable reception of multiple transmitted packets. Packet combining is commonly used to improve the performance of Hybrid-ARQ [120], and can be based on either hard decisions [169] or on soft (floating point) channel outputs [29].
A cooperative diversity PHY layer receiver architecture is shown in Figure 6.2. As in a conventional receiver, demodulation and sampling occurs in the DEMOD module, yielding a soft symbol output stream. What’s new is that the soft symbols of previously received packets are stored in a SAMPLE BUFFER and a COMBINER merges the stored packets with the newly received packet. The precise action of the combiner depends on the type of cooperation.
Fig. 6.2 A Cooperative Diversity Decoder Architecture. Thick arrows denote high-rate soft symbol interfaces.
When diversity is achieved through repetition coding, the new packet is simply a copy of a previously received packet and DEMOD performs maximal ratio combining (MRC) on the soft symbols of the packet copies. In this case, the SAMPLE BUFFER can store the soft symbols corresponding to a linear combination of past received packets, regardless of how many packet copies are received for a particular message. This applies to both the AF and DF orthogonal relaying repetition-coding schemes described in Section 5.4.2.
Observe that in Figure 6.2, there are four interfaces subject to the bandwidth expansion associated with passing soft symbols. The storage requirements of the sample buffer are subject to the same expansion factor. We further observe that a MAC protocol will result in multiplexing and asynchronous transmissions of multiple messages. Even a single cooperative link might carry traffic corresponding to multiple message streams. A cooperative node must maintain a SAMPLE BUFFER for each in-progress message communication.
Consider next the CF and incremental relaying strategies described in Section 5.4.2. These strategies are non-regenerative, as is AF, and the new packet may contain new coded symbols. Similarly, non-regenerative DF with the relay using an additional codebook would also result in new coded symbols. These approaches require the destination to store soft samples for all received packets for decoding. The storage requirements of the SAMPLE BUFFER thus increase linearly with the number of received packets for that message. Furthermore, the DECODER component becomes considerably more complex because decoding is based on multiple transmissions from multiple transmitters employing multiple codebooks.
By comparison, the non-orthogonal strategies of Section 5.4.3 appear to be simpler since a message results in a received signal at the destination over a continuous time interval. In this case, the conventional receiver of Figure 6.1 that resolves a codeword from a vector of soft samples should be sufficient. In particular, the high-speed "loop" associated with the COMBINER, DECODER, and SAMPLE BUFFER can be omitted. On the other hand, the structure of the received signal is altered considerably when a relay begins transmitting. In particular, the received signal consisting of the initial source transmission and the subsequent combined transmission of the source and relay is based on multiple transmitters and multiple codebooks and bears resemblance to the received signal of non-regenerative orthogonal relaying. In this case, the structure of the DECODER element will be similarly complex.
In any event, for both orthogonal and non-orthogonal relaying strategies, the DECODER processes a data structure of soft samples and makes bit decisions. If the CRC check is satisfied, the decoded bits pass to the higher layers and the soft samples in the SAMPLE BUFFER are discarded. If the CRC fails, the cooperative decoder stores the soft symbols in the SAMPLE BUFFER and waits for additional received packets. In the absence of a mechanism yielding additional packets, the soft samples are discarded.
Cooperative PHY Layer: Relay Node
We next examine the cooperative PHY layer for a relay node. A conventional store-and-forward relay node passes reliably received packets to the network layer. The network layer reads the packet header and determines whether the packet should pass to higher layers or whether it should be forwarded. If the packet is to be forwarded, it is passed (typically with a modified header) back down to the PHY layer transmitter. This approach extends naturally to DF relaying. The key observation is that all data packet transmissions are generated by the network layer.
The situation is less clear with AF or CF. Under AF, the relay transmits a vector of received soft samples. The transmission is presumably prefixed with an appropriately descriptive header. This sort of header information is normally generated at the network layer, so it would be consistent with layering practice for this vector of unreliable soft symbols to be passed to the network layer that then constructs a "soft symbol packet" for transmission. While this may sound logical, it incurs the substantial penalty of a high-rate soft symbol interface between the PHY and network layers.
Similar issues arise for CF where the PHY layer receives a transmission block b from which it deduces a quantization codebook index sb (see Section 4.2.4). If the relay-destination channel is very good, the quantized observation may closely approximate the received signal. In this case, forwarding even the index to the network layer may result in a bandwidth expansion of the PHY-network layer interface comparable to that for AF. Avoiding this potential bandwidth expansion in the PHY-network layer interface is the chief reason to embed some network-layer functionality in the PHY layer.
